Skip to content
Weights Not Walls
Menu

microsoft

phi-2

Reviewed
Size
2.8B parameterscounted from the safetensors header on the Hub
Licence
custom, read it, MIT LicenseOur reading, check the licence: A custom licence we have not read for this model: other makers use the same name for different texts. Read the model's own licence file before you use it.Licence text
On the Hub since
13 Dec 2023the date the repo was created
Context length
2k tokensfrom the model's config
Kind
Base model (no base model declared on the Hub)PhiForCausalLM
On Hugging Face
microsoft/phi-2175k downloads in the last 30 days, 3.5k likes

Scores

10 independent results on 8 eval sets
ANLIReported

Epoch's name: phi-2. Best closed model on this set: gpt-3.5-turbo-1106, 58.1% (OpenAI). All 15 results

ARC AI2Reported

Epoch's name: phi-2. Best closed model on this set: gpt-3.5-turbo-1106, 87.4% (OpenAI). All 134 results

BBHReported

Epoch's name: phi-2. Best closed model on this set: gemini-1.5-pro-001, 89.2% (Google DeepMind). All 88 results

HellaSwagReported

Epoch's name: phi-2. Best closed model on this set: gpt-4-0314, 95.3% (OpenAI). All 113 results

MMLUReported

Epoch's name: phi-2. Best closed model on this set: gpt-4o-2024-11-20, 88.1% (OpenAI). All 217 results

OpenBookQAReported

Epoch's name: phi-2. Best closed model on this set: gpt-3.5-turbo-1106, 86.0% (OpenAI). All 70 results

TriviaQAReported

Epoch's name: phi-2. Best closed model on this set: claude-2.0, 87.5% (Anthropic). All 115 results

WinograndeReported

Epoch's name: phi-2. Best closed model on this set: claude-3-opus-20240229, 88.5% (Anthropic). All 137 results

Claimed by the maker Claimed

The model card has no claimed scores in a machine-readable form (the card's model-index), so none are shown. The maker may give scores in the card text: read the card.

Other views, which we link to and do not copy: look this model up on Artificial Analysis and the Arena leaderboard.

Independent scores from Epoch AI (CC-BY 4.0), fetched 19 Nov 2026: Epoch AI, ‘Capabilities & benchmarking’. Published online at epoch.ai. Retrieved from ‘https://epoch.ai/benchmarks’ [online resource]. Each row links its own source. What the badges mean.

Run it yourself

file sizes read from the Hub 19 Nov 2026

Memory needed = weights + context cache + 0.50 GB for the runtime. Context cache at 8k tokens: 2.5 GB, worked out from the model's config (microsoft/phi-2), at 16-bit precision. Choose a machine above to see a verdict for each format. Method

GGUF for llama.cpp, Ollama, LM Studio and others

TheBloke/phi-2-GGUF (converted by TheBloke), 19k downloads. Sizes are the real files on the Hub, all parts added up.

FormatSizeMemory neededGet it
Q2_K1.1 GBfile size4.1 GB1.1 + 2.5 + 0.50
Commands
llama.cpp
llama-server -hf TheBloke/phi-2-GGUF:Q2_K
Ollama
ollama run hf.co/TheBloke/phi-2-GGUF:Q2_K
Download
hf download TheBloke/phi-2-GGUF --include "phi-2.Q2_K.gguf" --local-dir phi-2-GGUF
Q3_K_S1.2 GBfile size4.2 GB1.2 + 2.5 + 0.50
Commands
llama.cpp
llama-server -hf TheBloke/phi-2-GGUF:Q3_K_S
Ollama
ollama run hf.co/TheBloke/phi-2-GGUF:Q3_K_S
Download
hf download TheBloke/phi-2-GGUF --include "phi-2.Q3_K_S.gguf" --local-dir phi-2-GGUF
Q3_K_M1.4 GBfile size4.4 GB1.4 + 2.5 + 0.50
Commands
llama.cpp
llama-server -hf TheBloke/phi-2-GGUF:Q3_K_M
Ollama
ollama run hf.co/TheBloke/phi-2-GGUF:Q3_K_M
Download
hf download TheBloke/phi-2-GGUF --include "phi-2.Q3_K_M.gguf" --local-dir phi-2-GGUF
Q4_01.5 GBfile size4.5 GB1.5 + 2.5 + 0.50
Commands
llama.cpp
llama-server -hf TheBloke/phi-2-GGUF:Q4_0
Ollama
ollama run hf.co/TheBloke/phi-2-GGUF:Q4_0
Download
hf download TheBloke/phi-2-GGUF --include "phi-2.Q4_0.gguf" --local-dir phi-2-GGUF
Q3_K_L1.5 GBfile size4.5 GB1.5 + 2.5 + 0.50
Commands
llama.cpp
llama-server -hf TheBloke/phi-2-GGUF:Q3_K_L
Ollama
ollama run hf.co/TheBloke/phi-2-GGUF:Q3_K_L
Download
hf download TheBloke/phi-2-GGUF --include "phi-2.Q3_K_L.gguf" --local-dir phi-2-GGUF
Q4_K_S1.5 GBfile size4.5 GB1.5 + 2.5 + 0.50
Commands
llama.cpp
llama-server -hf TheBloke/phi-2-GGUF:Q4_K_S
Ollama
ollama run hf.co/TheBloke/phi-2-GGUF:Q4_K_S
Download
hf download TheBloke/phi-2-GGUF --include "phi-2.Q4_K_S.gguf" --local-dir phi-2-GGUF
Q4_K_M1.7 GBfile size4.7 GB1.7 + 2.5 + 0.50
Commands
llama.cpp
llama-server -hf TheBloke/phi-2-GGUF:Q4_K_M
Ollama
ollama run hf.co/TheBloke/phi-2-GGUF:Q4_K_M
Download
hf download TheBloke/phi-2-GGUF --include "phi-2.Q4_K_M.gguf" --local-dir phi-2-GGUF
Q5_01.8 GBfile size4.8 GB1.8 + 2.5 + 0.50
Commands
llama.cpp
llama-server -hf TheBloke/phi-2-GGUF:Q5_0
Ollama
ollama run hf.co/TheBloke/phi-2-GGUF:Q5_0
Download
hf download TheBloke/phi-2-GGUF --include "phi-2.Q5_0.gguf" --local-dir phi-2-GGUF
Q5_K_S1.8 GBfile size4.8 GB1.8 + 2.5 + 0.50
Commands
llama.cpp
llama-server -hf TheBloke/phi-2-GGUF:Q5_K_S
Ollama
ollama run hf.co/TheBloke/phi-2-GGUF:Q5_K_S
Download
hf download TheBloke/phi-2-GGUF --include "phi-2.Q5_K_S.gguf" --local-dir phi-2-GGUF
Q5_K_M1.9 GBfile size4.9 GB1.9 + 2.5 + 0.50
Commands
llama.cpp
llama-server -hf TheBloke/phi-2-GGUF:Q5_K_M
Ollama
ollama run hf.co/TheBloke/phi-2-GGUF:Q5_K_M
Download
hf download TheBloke/phi-2-GGUF --include "phi-2.Q5_K_M.gguf" --local-dir phi-2-GGUF
Q6_K2.1 GBfile size5.1 GB2.1 + 2.5 + 0.50
Commands
llama.cpp
llama-server -hf TheBloke/phi-2-GGUF:Q6_K
Ollama
ollama run hf.co/TheBloke/phi-2-GGUF:Q6_K
Download
hf download TheBloke/phi-2-GGUF --include "phi-2.Q6_K.gguf" --local-dir phi-2-GGUF
Q8_02.8 GBfile size5.8 GB2.8 + 2.5 + 0.50
Commands
llama.cpp
llama-server -hf TheBloke/phi-2-GGUF:Q8_0
Ollama
ollama run hf.co/TheBloke/phi-2-GGUF:Q8_0
Download
hf download TheBloke/phi-2-GGUF --include "phi-2.Q8_0.gguf" --local-dir phi-2-GGUF

MLX for Apple silicon

FormatSizeMemory neededRepo
MLX 4bitjesusoctavioas/phi-2-mlx-4Bit1.5 GBfile size4.5 GB1.5 + 2.5 + 0.50Download page
MLX 4bitSerhiiBzenko/phi-2-mlx-4Bit1.5 GBfile size4.5 GB1.5 + 2.5 + 0.50Download page

Original weights size from the safetensors header, for vLLM, SGLang, Transformers

FormatSizeMemory neededRepo
F16original weights5.2 GBfile size8.2 GB5.2 + 2.5 + 0.50
hf download microsoft/phi-2 --local-dir phi-2
Other conversions (12; sizes not read)

The Hub tags 64 quantised repos as conversions of this model; we read the most downloaded and show the maker's own first.

Run it by API

Hugging Face Inference Providers, fetched 19 Nov 2026, 13:58 UTC

No provider on Hugging Face Inference Providers serves this model (as of 19 Nov 2026, 13:58 UTC). Other hosts may; we only list what the router reports.

Not built yet: one key for every open model, with BlockRun, and our own tested-quality score per provider. Until then, precision and measured speed are the best guide to a provider here.

Planned for this page Not built yet

Try it in private chat
A link to Solana AI's confidential chat for the models it serves. This model is not listed as served there.
Fine-tune it
Fine-tuning a few base models from this page and publishing the result here.
Creators and earnings
Pages for people who publish fine-tunes and adapters of this model, and what they earn.
Fund and bounties
Bounties for new open models, paid from a public fund. How the fund will work.
Contribute compute
How to give compute to open model work, with Akash.

Model data from the Hugging Face Hub 19 Nov 2026, 13:43 UTC. The repo. Found a mistake on this page? Tell us.